This one is kind of obvious - because people prompt coding LLMs with natural language. That's unrelated to stuffing the pre-train set with trivia factoids. An

This one is kind of obvious - because people prompt coding LLMs with natural language. That's unrelated to stuffing the pre-train set with trivia factoids. An LLM that knows English very well isn't actually very large and certainly not hundreds of billions of parameters.